New Stanford and Together AI paper finds that cheaper reasoning need not cost accuracy, as hybrid agents cut decoding compute by 62.5% to 72.9% and still scored higher. Shows that agents run faster and score higher when most thinking happens in hidden vectors, not written text.
Hybrid Agents Cut Decoding Compute 72.9%, Score Higher
A Stanford and Together AI paper finds hybrid agents cut decoding compute by 62.5%–72.9% while scoring higher, suggesting most thinking works best in hidden vectors, not written text.